TypeSafe

Overview

TypeSafe is a specialized decision provider serving the jev family of System One judgment models. A decision request carries a state (what to evaluate) and a map of named questions; the response carries one typed answer per question. Bifrost performs conversions including:

Model ID mapping - Uses provider model identifier directly (jev-1.13.0, jev-latest, jev-preview)
Question conversion - Maps Bifrost decision kinds (noul, choice, score) to TypeSafe’s native System One question types 1:1
Answer normalization - Native noul/choice/score value fields normalize to a unified value, with confidence, probabilities, and legend metadata preserved
Usage normalization - input_tokens/output_tokens map to Bifrost’s prompt_tokens/completion_tokens (TypeSafe bills input tokens only)
Strict validation - Unsupported kinds, missing instructions, and malformed criteria are rejected locally with a 400 rather than silently approximated

Supported Operations

Operation	Non-Streaming	Streaming	Endpoint
Decisions	✅	-	`/v1/decisions`
List Models	✅	-	`/v1/models`
Chat Completions	❌	❌	-
Responses API	❌	❌	-
Text Completions	❌	❌	-
Embeddings	❌	❌	-

Unsupported Operations (❌): Chat Completions, Responses API, Text Completions, Embeddings, and every other operation return UnsupportedOperationError - TypeSafe serves judgment models only.Model listing: TypeSafe documents no upstream models endpoint. Bifrost serves the jev catalog from its model datasheet, including pricing and context length.

Setup & Configuration

Configure TypeSafe as a provider with a bearer API key:

{
  "providers": {
    "typesafe": {
      "keys": [
        {
          "name": "TypeSafe API Key",
          "value": "env.TYPESAFE_API_KEY",
          "weight": 1,
          "models": ["*"]
        }
      ]
    }
  }
}

Decisions API

POST /v1/decisions evaluates state against named questions. Each question has a kind, instructions (string or structured data), and kind-specific criteria:

Kind	Answer value	Criteria
`noul`	Probability between 0 and 1	Optional `true`/`false` descriptions
`choice`	One option string	Required map of option to description (max 255)
`score`	Numeric rubric score, including fractions	Required ordered array of 2-10 level descriptions

curl http://localhost:8080/v1/decisions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "typesafe/jev-1.13.0",
    "state": "Customer message: I was double charged and nobody replied. I want a refund today.",
    "questions": {
      "is_frustrated": { "kind": "noul", "instructions": "Is the customer frustrated?" },
      "category": {
        "kind": "choice",
        "instructions": "Pick the ticket category",
        "criteria": { "billing": "charges and refunds", "bug": "product defects", "other": "anything else" }
      },
      "urgency": {
        "kind": "score",
        "instructions": "Rate how urgently this needs a human reply",
        "criteria": ["can wait a week", "should be answered soon", "needs a reply today"]
      }
    }
  }'

Response:

{
  "model": "jev-1.13.0",
  "answers": {
    "is_frustrated": { "kind": "noul", "value": 0.98 },
    "category": {
      "kind": "choice",
      "value": "billing",
      "confidence": 1,
      "probabilities": { "billing": 1, "bug": 0, "other": 0 }
    },
    "urgency": {
      "kind": "score",
      "value": 2,
      "confidence": 1,
      "probabilities": { "0": 0, "1": 0, "2": 1 },
      "legend": { "0": "can wait a week", "1": "should be answered soon", "2": "needs a reply today" }
    }
  },
  "usage": { "prompt_tokens": 437, "completion_tokens": 72, "total_tokens": 509 }
}

Field Mapping Reference

Complete mapping between Bifrost’s /v1/decisions contract and TypeSafe’s native /v1/systemone API.

Request

Bifrost field	Native field	Accepted shapes	Validation
`model`	`model`	`typesafe/<id>` (routing prefix stripped)	Must resolve to the typesafe provider
`state`	`state`	string, object, array (lossless)	Any other shape (number, boolean, null) rejected with 400
`questions`	`questions`	Map of 1+ named questions	Empty map rejected with 400
`questions.<name>.kind`	`questions.<name>.type`	`noul`, `choice`, `score`	Any other kind rejected with 400
`questions.<name>.instructions`	`instructions`	string, object, array (lossless)	Required; other shapes rejected with 400
`questions.<name>.criteria`	`criteria`	Kind-specific, see below	Kind-specific, see below
`fallbacks`	-	`["provider/model", ...]`	Bifrost-only; consumed by fallback routing, never sent upstream

Criteria by kind

Kind	Bifrost / native shape (identical)	Validation
`noul`	Optional map with only `"true"` and `"false"` keys, string descriptions	Other keys or non-string descriptions rejected with 400
`choice`	Required map of option → description (string)	Missing, empty, non-map, non-string descriptions, or more than 255 options rejected with 400
`score`	Required ordered array of level descriptions	Fewer than 2 or more than 10 levels, or any non-string level, rejected with 400

In the converted request, validated criteria values pass through to TypeSafe unchanged - descriptions are never rewritten. The conversion path re-encodes JSON, so whitespace and object-key ordering can differ; raw-request passthrough bypasses conversion and preserves the original request bytes.

Response

Bifrost field	Built from native	Notes
`model`	`model`	Aliases resolve upstream: request `jev-latest`, response reports `jev-1.13.0`
`answers.<name>.kind`	`answers.<name>.type`	Identical vocabulary (`noul`/`choice`/`score`)
`answers.<name>.value`	`answers.<name>.noul` \| `.choice` \| `.score`	Unified value field: number in [0,1] for noul (out-of-range provider values rejected), option string for choice, fractional rubric score for score
`answers.<name>.confidence`	`confidence`	Present on choice/score answers when supplied
`answers.<name>.probabilities`	`probabilities`	Distribution over options (choice) or level indices (score)
`answers.<name>.legend`	`legend`	Level index → description map for score answers
`usage.prompt_tokens`	`usage.input_tokens`	TypeSafe bills input tokens only
`usage.completion_tokens`	`usage.output_tokens`	Output rate is $0
`usage.total_tokens`	-	Computed: input + output
`extra_fields`	-	Bifrost-only: request type, routing info, latency, raw request/response capture

Every requested question must produce an answer of its declared kind; a missing answer, a kind mismatch, or a missing value fails the request rather than returning partial results.

Errors

Upstream status	Meaning	Bifrost behavior
401	Invalid or missing API key	Status and message preserved; key marked failed, other keys rotate in
422	Request validation failed upstream	Status and validation detail preserved
429	Rate limited	Retried with backoff; keys rotate (per-key limit)
529	Service overloaded	Retried with backoff on the same key (capacity, not credential)

Requests rejected by Bifrost’s own validation (see tables above) return 400 with caller_invalid_request before any upstream call is made.

Native Integration

Bifrost also exposes TypeSafe’s native API 1:1 under the /typesafe prefix, for clients written against TypeSafe directly:

Native Endpoint	Method	Notes
`/typesafe/v1/systemone`	POST	Native request/response shape; questions use `type` instead of `kind`
`/typesafe/v1/models`	GET	Native `{"models": [{"name", "description", "release_date"}]}` shape, served from the datasheet

The native route accepts bare model IDs (jev-1.13.0) exactly as TypeSafe does, and additionally accepts typesafe/-prefixed IDs.

Success responses are shape-compatible with TypeSafe’s own API - identical byte-for-byte only on the raw-response passthrough path, otherwise rebuilt and re-encoded (native shape preserved, exact JSON byte ordering not guaranteed). Error responses use TypeSafe’s native {"detail": {"error_type", "message"}} shape with upstream status codes preserved. Requests rejected by Bifrost’s local validation return 400 where TypeSafe’s own validation would return 422.

Model Naming

Model	Notes
`jev-1.13.0`	Current versioned release
`jev-latest`	Alias, currently resolves to `jev-1.13.0`
`jev-preview`	Preview alias, currently resolves to `jev-1.13.0`

Responses always report the resolved versioned model. TypeSafe bills input tokens only; the jev pricing, 64k context window, and rate limits are documented at docs.typesafe.ai/models.

Overview

Quick Start

Release Cadence

Migration Guides

Migrate from other Gateway

SDK Integrations

Providers & Guides

MCP Gateway

Custom plugins

Open Source Features

Overview

Supported Operations

Setup & Configuration

Decisions API

Field Mapping Reference

Request

Criteria by kind

Response

Errors

Native Integration

Model Naming

​Overview

​Supported Operations

​Setup & Configuration

​Decisions API

​Field Mapping Reference

​Request

​Criteria by kind

​Response

​Errors

​Native Integration

​Model Naming

Overview

Supported Operations

Setup & Configuration

Decisions API

Field Mapping Reference

Request

Criteria by kind

Response

Errors

Native Integration

Model Naming